I've been running kfold cross validation with 10 folds and comparing it against a test set. Logging the score and the stdev along the way.
Once I wad happy with my model I then run the estimator through the cross validation with the full data (including the test set) but ... The standard deviation of the scores increases and the score decreased????
Is this normal? I assumed MORE data decreases variance ? Or is it just due to the randomness in the splits ?